The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructions promote the use of npx to download and execute remote packages at runtime, specifically the mcporter tool and various MCP servers.
Evidence: npx mcporter list and npx -y @modelcontextprotocol/server-filesystem.
[COMMAND_EXECUTION]: The skill provides patterns for executing shell commands and scripts via the --stdio flag in both the list and call subcommands.
Evidence: mcporter list --stdio "npx -y @modelcontextprotocol/server-filesystem" and mcporter call --stdio "bun run ./server.ts".
[DYNAMIC_EXECUTION]: The generate-cli and emit-ts commands generate executable code or CLI wrappers based on remote server configurations, which are then stored or executed on the host system.
Evidence: mcporter generate-cli --server <name> and mcporter emit-ts <server> --mode client.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external MCP servers and tools, creating a vulnerability surface where malicious external content could influence agent behavior.
Ingestion points: External tool outputs retrieved via mcporter call <server.tool> (found in SKILL.md).
Boundary markers: None identified; instructions do not specify the use of delimiters or sanitization wrappers for tool outputs.
Capability inventory: The skill has the ability to execute shell commands (--stdio), write files (generate-cli), and manage authentication configurations (mcporter auth).
Sanitization: No evidence of input validation or output escaping is present in the skill instructions.
[EXTERNAL_DOWNLOADS]: The skill encourages the installation and use of mcporter, a community-authored package from the public npm registry, as well as connections to arbitrary HTTP MCP servers.
Evidence: npm install -g mcporter and mcporter list --http-url https://some-mcp-server.com.

mcporter